8tracks

Discovery Logic for Playlists based on the tags.

How to Run

1. clone the repository.

2. create the virtual environment using command - virtualenv <name of the environment>

3. Activate the virtual environment using command - source <name of the environment>/bin/activate

4. Install all the required packages by running command - pip install -r requirements.txt

5. I tried dumping and undumping data in .sql file but got some errors. So one must have to do some work to give some data to the system::

   a) create a database named "tracks". b) Go to the settings file in the project and update your mysql user and password.

   DATABASES = {
   'default': {
       'ENGINE': 'django.db.backends.mysql',
       'NAME': 'tracks',
       'USER': 'root',
       'PASSWORD': 'alpha123',
       'HOST': 'localhost',
       'PORT': '',
   },
   }
   

   c) After updating mysql user and password sync the databases using command - python manage.py migrate

6. Run the server by managemant command - python manage.py runserver

7. Now it's time to populate the database. I have created two extra APIs to post playlists and tags directly from the browser::

   For Adding new PlayLists hit the API - localhost:8000/api/playlists/playlists/ This API will list all the playlists in the system and one can also add new playlists into the system. For adding new tags hit the API - localhost:8000/api/playlists/tags/ This API will list all the tags in the system and one can also add new tags into the system.

8. I have kept the relationship between Tags and PlayLists as flexible. I did not define an end-point API for creating relationships between these. This can be done by inserting some entries into the playlists_playlist_tags in the tracks database as this table contains the entries for many to many relationship between Playlists and tags.

9. IMPORTANT - Every time you add new Tag instances, you have to push Tag indices to the AWS instance by using management command - python manage.py update_index

10. For Search API - hit the URL http://localhost:8000/api/search/playlists/?q=<tag_name1><tag_name2>. For Ex. - http://localhost:8000/api/search/playlists/?q=poprock. here pop and rock are the two tag_names. This API will return the lists of playlists based on relevancy and sorted on basis of configuration 'ORGANIZE_BY' parameter in settings.py.

Detail of any PlayList can also be viewed via URL - http://localhost:8000/api/playlists/playlists/<playlist_id>/ where playlist_id can be unique id of any playlist in the system.

For ex - http://localhost:8000/api/playlists/playlists/1/ returned on my system -

HTTP 200 OK
Allow: GET, PUT, PATCH, HEAD, OPTIONS
Content - Type: application / json
Vary: Accept

{
    "playlist_id": 1,
    "playlist_name": "Led Zeppelin",
    "likes": 10,
    "plays": 3
}

Similarily Detail of any Tag can also be viewed via URL - http://localhost:8000/api/playlists/playlists/<tag_id>/ where tag_id can be unique_id of any Tag in the system.

Methodology Used -

I have used the concept of Elastic Search and exposing APIs using django-rest-framework. Elastic search is used to do fast searching.

I have used a AWS Elastic search instance as a remote server for hosting elastic search. The index name is haystack and it stores the indices of Tag model. You can thus do a query search on thousand of tags.